Order reduction method for 3D steady state model based on SVD-Kriging

In response to the issues of high computational load and low efficiency in the simulation of three-dimensional steady-state models for engineering machinery, a model order reduction method based on the SVD-Kriging approach (singular value decomposition—Kriging method) is proposed. First, a dataset of the three-dimensional steady-state system is obtained through simulation calculations. The output matrix is decomposed based on SVD-LOO (Leave-One-Out) to obtain the optimal order and corresponding basis functions; secondly, by using the Kriging model and basis functions, a mapping relationship between input parameters and the output matrix is constructed, ultimately achieving multi-point steady-state responses of the three-dimensional system by providing any set of steady-state input constants. The validation results of this method on the three-dimensional steady-state pipeline system dataset indicate that this order reduction method can effectively improve computational efficiency while maintaining high prediction accuracy. In addition, using this method for order reduction calculations of the temperature field in the electric front crane battery system and applying it to digital twin technology to assist in optimizing the energy consumption of the electric front crane further demonstrates the effectiveness and practicality of this method in real engineering applications.